nfarrukh7

Clinical Informatics Fellow at Nationwide and The Ohio State. Hospitalist. Cant forget - YNWA!

[nfarrukh7 @nfarrukh7]

@maria_rcks @sama Realistically if the issue is compute, decreasing the friction it takes to help me choose a less compute-intensive model to get my action done instead of defaulting to having Sol on all the time would go such a long way

[nfarrukh7 @nfarrukh7]

@ericrovner @AlexFinn This is quite astute. I think Dennis Hasabi or Jensen popularized this term, capability overhang, which describes how the current models still aren't being used by most people to the top of their capabilities.

[nfarrukh7 @nfarrukh7]

Have been sharing with all my informatics connected friends. Brought a campaign on my wife (whose in medicine too) to listen to this.

[Christopher A. Longhurst, MD @CALonghurst]

Epic has 607 health system customers with ~3200 hospitals between them, representing ~590k physicians and 325M unique patients (285M in the US). 58% of US doctors use Epic. Want to learn more? This deep dive is long but worth the listen 🎧 ⤵️ @HeyEpic

https://t.co/m34ogWURBn

[Christopher A. Longhurst, MD @CALonghurst]

𝐌𝐞𝐚𝐬𝐮𝐫𝐢𝐧𝐠 𝐀𝐈 𝐚𝐠𝐚𝐢𝐧𝐬𝐭 𝐭𝐡𝐞 𝐇𝐞𝐚𝐥𝐭𝐡 𝐂𝐚𝐫𝐞 𝐖𝐞 𝐇𝐚𝐯𝐞

"Investigators should be careful to compare outcomes associated with the use of AI with outcomes in the health care systems we have, rather than the ones we wish we had."

💯

[nfarrukh7 @nfarrukh7]

@AdanBecerraPhD @jonc101x @emollick I'd argue it has to meet a certain threshold level to merit studying these outcomes, much the same way we test medical students and residents before sending them off to a job being evaluated on mortality, LOS, readmission.

[nfarrukh7 @nfarrukh7]

Anyone working in Healthcare will realize there are big and small gaps that providers and patients need help plugging. We are not close to replacement, who knows maybe one day? But imo those talking about replacement are ignoring the equity part of the equation.

[nfarrukh7 @nfarrukh7]

thanks for the courage of your voice as a part creator of this dataset to highlight some nuances. Sometimes it frustrates me that people take already incredible achievements by these models and take them out of context. The accuracy improvement stands alone as impressive.

[Cyril Zakka, MD @cyrilzakka]

I mean no disrespect to any of the authors in this thread but the claims being made are wild and/or taken out of context.
The AgentClinic dataset is based on an online dataset (read contamination) I helped create, from a resource aimed at helping medical students study for the Step 1 exam.
For the uninitiated, the Step 1 exam is aimed at making sure pre-clinical medical students have passing knowledge of medical concepts. It is in no way a representation of any real clinical workflows. I also reviewed the AgentClinic dataset and no, we’re not even close to replacing physicians. Please let’s taper these claims. They do more harm than good.

[nfarrukh7 @nfarrukh7]

@LindseyKnake @culehmann @pediatric_cds The ribbon game at Amia is an unspoken treasure.

[nfarrukh7 @nfarrukh7]

Wanted to thank these incredible mentors at Nationwide Childrens Hospital for giving me the opportunity to present on the Physician Informatics Consult Service and be part of a fun panel at #AMIA CIC 24

[Zaven Sargsyan @sargsyanz]

Does pancakeopenia mean there’s a deficiency of pancake, or a deficiency of all cakes

[nfarrukh7 @nfarrukh7]

This is a fantastic investment in the system. It is our role as physicians to step up deliver in return. Im going to read up on the genesis and development of this policy, so if anyone has good resources please send my way.

[Craig Joseph, MD @CraigJoseph]

Kentucky got this one right: "Kentucky first state to decriminalize medical errors." With exceptions for gross negligence or intentional acts, this law will promote a culture of sharing and learning from mistakes, much like with aviation. @Nordic_Global https://t.co/yc4Uwqa3vI

[nfarrukh7 @nfarrukh7]

Excited to be at my first #XGM #Epic

[Melina Manolas, MD, MPH @melina_manolas]

Don’t get me wrong I don’t miss residency but there’s something about that 7pm coffee on the way into a night shift that hits different

[nfarrukh7 @nfarrukh7]

@Dr_Oubre @dinaioffeMD Sage advice that's simple - actually write an assessment. It's your clinical reasoning on why it's this diagnosis vs. Others. Also what are the implications (the general/overarching approach of your plan). Preferably in paragraph then dashes for the plan (i.e. your to-dos)

[nfarrukh7 @nfarrukh7]

@Dr_Oubre Always a great discussion. Too much of my medical education was spent trying to craft these perfect notes and HPI, but now most of those sections are an after thought or are pulled in in a reasonable way via epic smartlinks. I can't say I was taught how to write a plan until...

[NEJM AI @NEJM_AI]

In a talk @BrownMedicine, Editor-in-Chief @zakkohane, urged doctors-in-training to understand and make use of artificial intelligence.

“It's an existential opportunity and threat to medicine to not take [AI] seriously.”

See what else he said: https://t.co/UHX9pTOnLU

[Ashish K. Jha @ashishkjha]

Ozempic (and drugs like it) will likely have a profound impact on health and our society at large

@DhruvKhullar explains in a really good @NewYorker piece

The term game-changer gets thrown around too lightly

But

These drugs might be game-changers

https://t.co/hK06QmhoDE

[nfarrukh7 @nfarrukh7]

@Dr_Oubre @flatley_bill Unfortunately one of the problems is the way echo "data" is handled. The EF value, depending on your IT infrastructure setup might not be a discrete value so it is difficult for Epic to pull in, but not out of the realm of possibility to display or even the whole report text.

[nfarrukh7 @nfarrukh7]

@Dr_Oubre @flatley_bill They have been making massive ground breaking improvements, but sometimes the clinician to developer pipeline doesn't connect and the ideas we have from front line experience don't trickle to the developers. I didn't see that idea in epic ideas:

https://t.co/sKwx4Vyo1R